An implementation of GraphQL for the crystal programming language inspired by graphql-ruby & go-graphql.
The library is in beta state atm. Should already be usable but expect to find bugs (and open issues about them). pull-requests, suggestions & criticism are very welcome!
Find the api docs here.
Add this to your application's shard.yml:
dependencies:
graphql-crystal:
github: ziprandom/graphql-crystalComplete source here.
Given this simple domain model of users and posts
class User
property name
def initialize(@name : String); end
end
class Post
property :title, :body, :author
def initialize(@title : String, @body : String, @author : User); end
end
POSTS = [] of Post
USERS = [User.new("Alice"), User.new("Bob")]We can instantiate a GraphQL schema directly from a graphql schema definition string
schema = GraphQL::Schema.from_schema(
%{
schema {
query: QueryType,
mutation: MutationType
}
type QueryType {
posts: [PostType]
users: [UserType]
user(name: String!): User
}
type MutationType {
post(post: PostInput) : PostType
}
input PostInput {
author: String!
title: String!
body: String!
}
type UserType {
name: String
posts: [PostType]
}
type PostType {
author: UserType
title: String
body: String
}
}
)Then we create the backing types by including the GraphQL::ObjectType and defining the fields using the field macro
# reopening User and Post class
class User
include GraphQL::ObjectType
# defaults to the method of
# the same name without block
field :name
field :posts do
POSTS.select &.author.==(self)
end
end
class Post
include GraphQL::ObjectType
field :title
field :body
field :author
endNow we define the top level queries
# extend self when using a module or a class (not an instance)
# as the actual Object
module QueryType
include GraphQL::ObjectType
extend self
field :users do
USERS
end
field :user do |args|
USERS.find( &.name.==(args["name"].as(String)) ) || raise "no user by that name"
end
field :posts do
POSTS
end
end
module MutationType
include GraphQL::ObjectType
extend self
field :post do |args|
user = USERS.find &.name.==(
args["post"].as(Hash)["author"].as(String)
)
raise "author doesn't exist" unless user
(
POSTS << Post.new(
args["post"].as(Hash)["title"].as(String),
args["post"].as(Hash)["body"].as(String),
user
)
).last
end
endFinally set the top level Object Types on the schema
schema.query_resolver = QueryType
schema.mutation_resolver = MutationTypeAnd we are ready to run some tests
describe "my graphql schema" do
it "does queries" do
schema.execute("{ users { name posts } }")
.should eq ({
"data" => {
"users" => [
{
"name" => "Alice",
"posts" => [] of String
},
{
"name" => "Bob",
"posts" => [] of String
}
]
}
})
end
it "does mutations" do
mutation_string = %{
mutation post($post: PostInput) {
post(post: $post) {
author {
name
posts { title }
}
title
body
}
}
}
payload = {
"post" => {
"author" => "Alice",
"title" => "the long and windy road",
"body" => "that leads to your door"
}
}
schema.execute(mutation_string, payload)
.should eq ({
"data" => {
"post" => {
"title" => "the long and windy road",
"body" => "that leads to your door",
"author" => {
"name" => "Alice",
"posts" => [
{
"title" => "the long and windy road"
}
]
}
}
}
})
end
endTo ease working with input parameters custom structs can be registered to be instantiated from the json params of query and mutation requests. Given the schema from above one can define a PostInput struct as follows
struct PostInput < GraphQL::Schema::InputType
JSON.mapping(
author: String,
title: String,
body: String
)
endand register it in the schema like:
schema.add_input_type("PostInput", PostInput)Now the argument post which is expected to be a GraphQL InputType PostInput will be automatically parsed into a crystal PostInput-struct. Thus the code in the post mutation callback becomes more simple:
module MutationType
include GraphQL::ObjectType
extend self
field :post do |args|
input = args["post"].as(PostInput)
author = USERS.find &.name.==(input.author) ||
raise "author doesn't exist"
POSTS << Post.new(input.title, input.body, author)
POSTS.last
end
endCustom context types can be used to pass additional information to the object type's field resolves. An example can be found here.
A custom context type should inherit from GraphQL::Schema::Context and therefore be initialized with the served schema and a max_depth.
GraphQL::Schema::Schema#execute(query_string, query_arguments = nil, context = GraphQL::Schema::Context.new(self, max_depth))accepts a context type as its third argument.
Field resolver callbacks on object types (including top level query & mutation types) get called with the context as their second argument:
field :users do |args, context|
# casting to your custom type
# is necessary here
context = context.as(CustomContext)
unless context.authenticated
raise "Authentication Error"
end
...
endFor an example of how to serve a schema over a webserver(kemal) see kemal-graphql-example.
The parser has been implemented using my crystal language toolkit and is significantly slower than the c implementation for larger schema strings while performing ok on smaller query strings. See benchmark/compare_benchmarks.cr for the strings used in the test.
To compare the performance of the Parser with facebooks GraphQL parser you need to have the library installed on your machine. Then run
crystal build --release benchmark/compare_benchmarks.crSCHEMA String: c implementation from facebook: 64.87k ( 15.41µs) (± 1.89%) fastest
SCHEMA String: cltk based implementation: 1.4k (713.07µs) (± 8.29%) 46.26× slower
QUERY String: c implementation from facebook: 16.63k ( 60.13µs) (± 2.00%) fastest
QUERY String: cltk based implementation: 5.62k ( 178.0µs) (± 4.34%) 2.96× slowerrun tests with
crystal spec
- Fork it ( https://github.com/ziprandom/graphql-crystal/fork )
- Create your feature branch (git checkout -b my-new-feature)
- Commit your changes (git commit -am 'Add some feature')
- Push to the branch (git push origin my-new-feature)
- Create a new Pull Request
- ziprandom - creator, maintainer